VMworld EMC today demoed how its upcoming fully-automated storage tiering (FAST) technology working together with SSDs might drastically improve storage performance and efficiency.
At this week's VMworld conference in San Francisco, EMC's Chad Sakac, vp of the VMware Technology Alliance and the geek behind the blog Virtual Geek, hosted a session grippingly entitled "Infrastructure Architectures Purpose-Built for the Virtual Datacenter."
During a wide-ranging discussion of EMC/VMware hardware and software interaction, Sakac asked the rhetorical question: "What would it mean if storage could completely tier itself?"
And one of the key tiers in both his argument and his demo was fast - and currently expensive - SSDs.
"We've literally shipped and sold out of solid state disks - enterprise flash disks - for the last six quarters. This is something which is transforming the storage business and the price curve is on Moore's Law: those things can do a lot more for a lot less, and frankly over the next one-to-two years will actually come to dominate the performance segment of the storage race," Sakac said.
But no matter how Moorsian are SSD's price drops, SATA drives are becoming increasingly cost-effective as well. "The challenge is that no one's going to replace everything that they've got with [SSDs], so you need to be able to figure out how to move things around. And the flip side is also true: you can now buy a 2TB SATA drive, and talking with disk vendors, the 6TB SATA drives are around the corner."
As Sakac pointed out, EMC is first introducing tiering technology at the LUN and the file-system level and then at the sub-LUN file-system level.
To demonstrate a FAST-enabled tiered setup of SSDs and SATA drives, Sakac's demo first showed a red-hot heat map of a 384-disk storage array thrashing away with heavy SATA-disk usage.
Sakac logged into the array and set up a policy for the tiering - "kind of like a DRS [distributed route scheduling] policy," he explained, meaning a VMware feature that allows you to set up how your resources are shared among virtual machines.
His demo's policy replaced 16 of the SATA disks with high-speed SSDs. After assigning that policy to a pair of vSphere clusters with a simple drag-and-drop, the FAST utility automatically redistributed the load among the tiers to take advantage of the added SSDs.
Although Sakac didn't let the demo run for the 20 minutes that would be needed for the full automatic retiering, he displayed a series of screenshots of a previously run iteration that showed the toasty red array automatically redistributing its load to take advantage of the SSDs and becoming a soothing blue and green.
"We've gone from 80 to 90 per cent utilization to 10 to 15 per cent utilization, he said. "What that translates into is not only better performance, but that means that you can literally put four times more performance workloads on that same infrastructure. Or thinking about it the other way, you need four times less stuff to begin with."
Sakac told the attendees that EMC has "basically" committed to deliver FAST across all of its platforms and across all protocols. "Whether you're using this on NAS or you're using this on block [protocol]," he said, "whether you're a mid-range customer or you're a high-end customer, this is something that [you're] going to need."
EMC has previously said that FAST technology at the LUN level will appear near the end of this year and that the sub-LUN version will come out sometime in mid-2010.
The "A" in FAST - Automated - is part of EMC's goal of, as Sakac put it, becoming invisible to system administrators. Ending his talk, he said: "You may think it's weird for us as a storage vendor to go, 'We want to make our stuff invisible!' We do, because we think that the value in the future is actually [storage's] degree of invisibility."
We couldn't agree more. We look forward to the day when no one needs to know what LUN a VM instance is running on because that LUN will simply take care of itself - by becoming invisible. ®